Lesson 9 of 25

The Ladder of Logical Progression (Step-by-Step Thinking)

Start with the honest part: adding the phrase "think step by step" to a prompt does very little on current models, because they already reason before answering. What still works, and works well, is you deciding the stages - naming the parts of the task, their order, and what each must produce. That is not a trick to make the model think harder; it is a structure that makes its reasoning checkable by you and lets you reject one stage without throwing away the rest.

Reach for it whenWhen a task has parts that depend on each other, or when you need to be able to verify an answer rather than just receive it.

The numbered-stage structure in the original is sound and needs no change. Two things are worth adding: make one stage a grounded research stage, with sources and dates, since in business and planning problems that stage is usually what decides the answer; and send the staged output where it will be worked on - a Sheet with a tab per phase and a column for dependencies is a working document, where the same content in a chat is a wall of text.

Worth knowingThe phrase "think step by step" is not needed here either. And a Sheet with formulas is more trustworthy for arithmetic than text in the chat, because the formula recomputes while a typed number does not - so for calculations, ask for the formula rather than the result.
  • Make one stage grounded research, with dates and sources.
  • Ask for formulas in a Sheet, not typed-out results.
  • One tab per phase, plus a dependencies column.
  • Doc for an analysis others will comment on.
  • Keep caveats in chat, out of the deliverable.
Example 1: Solving a Logic/Math Problem

Instead of

A train leaves Hyderabad at 3 PM traveling at 60 km/h. A car leaves Hyderabad at 3:30 PM on the same route traveling at 80 km/h. At what time will the car overtake the train?

Perplexity version

Solve this word problem and show the working in a form I can check.

Problem: A train leaves Hyderabad at 3 PM travelling at 60 km/h. A car leaves Hyderabad at 3:30 PM on the same route travelling at 80 km/h. At what time will the car overtake the train?

Structure the answer this way:

1. **Setup** - what is known, what is asked, and anything ambiguous in the problem.
2. **Working** - a table with columns Quantity, Value, Unit, How derived. Rows for head start distance, relative speed, time to close the gap, overtake time.
3. **Answer** - one line.
4. **Independent check** - compute each vehicle's distance travelled at that time, working forward from its own departure, and confirm they match. Do not reuse the relative-speed figure; that checks the arithmetic against itself.
5. **Assumptions** - constant speeds, no stops, identical route.

Then, because this is a teaching example and not just a sum:

* Generate the same problem with five different sets of numbers, and put all six versions into a Google Sheet called "Overtake problems" with columns Problem, Head start, Relative speed, Answer, and a Working column. Make the Answer column a formula wherever the arithmetic allows it, so the Sheet itself checks the answers rather than me trusting the text.
* Then tell me which step students most commonly get wrong, and what the wrong answer looks like.

I have not asked you to "think step by step" - that phrase does nothing for you now. The structure above is the actual instruction.
Open Perplexity 1,502 characters
Example 2: Analyzing a Business Decision

Instead of

Is opening a second branch of my restaurant in Begumpet a good idea?

Perplexity version

Act as a business consultant. I run a successful biryani house in Kondapur, Hyderabad, and I am considering a second branch in Begumpet.

Structure this in stages, and ground the market stage in real data rather than impressions.

**Stage 1 - What you need from me.** Revenue and margin, how much of Kondapur's success rests on my own presence, kitchen capacity, staff depth, capital available. Ask; do not assume. If the gaps make the later stages guesswork, say so.

**Stage 2 - Grounded market research, cited, with dates.** Use Google Search for: commercial rental ranges for restaurant space in Begumpet against Kondapur; which biryani and multi-cuisine restaurants currently operate in Begumpet and how they are rated; the residential and office mix and any recent development or metro connectivity that changes footfall; what has been reported about restaurant failure rates and second-outlet expansion in Indian metros. Show the sources and dates - this stage is why you are doing this here rather than elsewhere.

**Stage 3 - The case for.** At least three benefits specific to Begumpet and to biryani.

**Stage 4 - The case against.** At least three risks, including the ones owners underestimate: attention splitting, quality drift in a second kitchen, cannibalisation, lease carry before breakeven.

**Stage 5 - Recommendation.** A position, not a shrug. Then the single most important thing to establish before deciding, and how to establish it in thirty days.

Put stages 2 to 5 into a Google Doc called "Begumpet branch - analysis" so I can share it with my partner and he can comment in the margins. Keep stage 1's questions in the chat where I will answer them.

Mark clearly which claims rest on my facts, which on cited sources, and which on your assumptions.
Open Perplexity 1,778 characters
Example 3: Planning a Complex Task

Instead of

Give me a plan to launch a podcast.

Perplexity version

I want to launch a podcast about the history of Indian startups, starting from nothing.

Build the plan as three phases with a decision gate between each - the one question that must be answered before the next phase starts. A flat checklist lets me buy equipment before deciding who the show is for.

**Phase 1 - Pre-production:** niche and audience; name and branding; format and episode length; equipment at entry, mid and serious tiers in rupees; hosting platform.
**Phase 2 - Production and post-production:** guest sourcing and outreach; recording setup including remote guests; editing workflow; show notes and transcripts; cadence and episodes banked before launch.
**Phase 3 - Launch and marketing:** directory submissions; launch sequence; distribution and clips; promotion over the first ten episodes; what to measure and what to ignore.

Ground the parts that depend on current facts, with sources and dates:

* Current hosting platform options and pricing, and which accept Indian payment methods without friction.
* Equipment actually available in India at each tier, with current rupee prices - not US prices converted.
* Which podcast platforms Indian listeners actually use, as reported.
* What existing podcasts already cover Indian startup history, so I can see whether the niche is occupied. Cite them.

Then build the deliverable as a Google Sheet called "Podcast launch plan" with tabs for the three phases, columns Step, Owner, Depends on, Status, Notes, and a Gates tab holding the three gate questions. I will work out of the Sheet for months - that is why it goes there and not into the chat.

In the chat, separately: the steps people most commonly do too early and what it costs them, and the two or three steps that genuinely cannot be skipped.

Note I have not asked you to think step by step. The phases and gates are the instruction; the phrase would add nothing.
Open Perplexity 1,895 characters